The NEORV32 RISC-V Processor - Datasheet

Most of the RTL sources use entity instantiation. Hence, the RTL compile order might be relevant (depending on the synthesis/simulation tool. Therefore, two file-list files are provided in the rtl folder that list all required HDL files for the CPU core and for the entire processor and also represent their recommended compile order. These file-list files can be consumed by EDA tools to simplify project setup.

A simple bash script generate_file_lists.sh is provided for regenerating the file-lists (using GHDL’s elaborate command). This script can also be invoked using the default application makefile (see Makefile Targets).

By default, the file-list files include a placeholder in the path of each included hardware source file. These placeholders need to be replaced by the actual path before being used. Example:

The single clock domain of the processor can be split into an always-on clock domain and a switchable clock domain. The switchable clock domain is used to clock the CPU core, the CPU’s bus switch and - if implemented - the caches. This domain can be deactivated to reduce power consumption. The always-on clock domain is used to clock all other processor modules like peripherals, memories and IO devices. Hence, these modules can continue operation (e.g. a timer keeps running) even if the CPU is shut down.

The splitting into two clock domain is enabled by the CLOCK_GATING_EN generic (Processor Top Entity - Generics). When enabled, a generic clock switching gate is added to decouple the switchable clock from the always-on clock domain (VHDL file neorv32_clockgate.vhd). Whenever the CPU enters Sleep Mode the CPU clock domain ist shut down.

Many processor modules like the UARTs or the timers provide a programmable time base for operations. In order to simplify the hardware, the processor implements a global "clock generator" (neorv32_sys.vhd) that provides single-cycle clock enables for certain frequencies which are derived from the main clock. These clock enable signals are synchronous to the system’s main clock. The processor modules can use these enables for sub-main-clock operations while still providing a single clock domain only.

In total, 8 sub-main-clock signals are available. All processor modules, which feature a time-based configuration, provide a programmable three-bit prescaler select in their control register to select one of the 8 available clocks. The mapping of the prescaler select bits to the according clock source is shown in the table below. Here, f represents the processor main clock from the top entity’s clk_i signal.

The processor setup features the standard machine-level RISC-V interrupt lines for "machine timer interrupt", "machine software interrupt" and "machine external interrupt". Their usage is defined by the RISC-V privileged architecture specifications. However, bare-metal system can also repurpose these interrupts. See CPU section Traps, Exceptions and Interrupts for more information.

As part of the NEORV32-specific CPU extensions, the processor core features 16 fast interrupt request signals (FIRQ0 to FIRQ15) providing dedicated bits in the mip and mie CSRs and custom mcause trap codes. The FIRQ signals are reserved for processor-internal modules only (for example for the communication interfaces to signal "available incoming data" or "ready to send new data").

The mapping of the 16 FIRQ channels to the according processor-internal modules is shown in the following table (the channel number also corresponds to the according FIRQ priority: 0 = highest, 15 = lowest):

The CPU can access all of the 32-bit address space from the instruction fetch interface and also from the data access interface. Both CPU interfaces can be equipped with optional caches (Processor-Internal Data Cache (dCACHE) and Processor-Internal Instruction Cache (iCACHE)). The two CPU interfaces are multiplexed by a simple bus switch into a single processor-internal bus. Optionally, this bus is further switched by another instance of the bus switch so the Direct Memory Access Controller (DMA) controller can also access the entire address space. Accesses via the resulting SoC bus are split by the Bus Gateway that redirects accesses to the according main address regions (see table above). Accesses to the processor-internal IO/peripheral devices are further redirected via a dedicated IO Switch.

The central bus gateway serves two purposes: redirect core accesses to the according modules (e.g. memory accesses vs. memory-mapped IO accesses) and monitor all bus transactions. The redirection of access request is based on a customizable memory map implemented via VHDL constants in the main package file (rtl/core/neorv323_package.vhd):

Besides the delegation of bus requests the gateway also implements a bus monitor (aka "the bus keeper") that tracks all active bus transactions to ensure safe and deterministic operations.

Whenever a memory-mapped device is accessed (a real memory, a memory-mapped IO or some processor-external module) the bus monitor starts an internal timer. The accessed module has to respond ("ACK") to the bus request within a specific time window. This time window is defined by a global constant in the processor’s VHDL package file (rtl/core/neorv323_package.vhd).

This constant defines the maximum number of cycles after which a non-responding bus request (i.e. no ack and no err signal) will time out raising a bus access fault exception. For example this can happen when accessing "address space holes" - addresses that are not mapped to any physical module. The resulting exception type corresponds to the according access type, i.e. instruction fetch access exception, load access exception or store access exception.

The reservation set controller is responsible for handling the load-reservate and store-conditional bus transaction that are triggered by the lr.w (LR) and sc.w (SC) instructions from the CPU’s Zalrsc ISA Extension.

A "reservation" defines an address or address range that provides a guarding mechanism to support atomic accesses. A new reservation is registered by the LR instruction. The address provided by this instruction defines the memory location that is now monitored for atomic accesses. The according SC instruction evaluates the state of this reservation. If the reservation is still valid the write access triggered by the SC instruction is finally executed and the instruction return a "success" state (rd = 0). If the reservation has been invalidated the SC instruction will not write to memory and will return a "failed" state (rd = 1).

The reservation is invalidated if…​

Bus transactions triggered by the LR instruction register a new reservation set and are delegated to the adressed memory/device. Bus transactions triggered by the SC remove a reservation set and are forwarded to the adressed memory/device only if the SC operations succeeds. Otherwise, the access request is not forwarded and a local ACK is generated to terminate the bus transaction.

The IO switch further decodes the address when accessing the processor-internal IO/peripheral devices and forwards the access request to the according module. Note that a total address space size of 256 bytes is assigned to each IO module in order to simplify address decoding. The IO-specific address map is also defined in the main VHDL package file (rtl/core/neorv323_package.vhd).

This is the most common and thus, the default boot configuration. When selected, the processor-internal Bootloader ROM (BOOTROM) is enabled. This ROM contains the executable image (rtl/core/neorv32_bootloader_image.vhd) of the default NEORV32 Bootloader that will be executed right after reset. The bootloader provides an interactive user console for executable upload as well as an automatic boot-configuration targeting external (SPI) memories.

If the processor-internal Instruction Memory (IMEM) is enabled it will be implemented as blank RAM.

This is the most flexible boot configuration as it allows the user to specify a custom boot address via the BOOT_ADDR_CUSTOM generic. Note that this address has to be aligned to 4-byte boundary. The processor will start executing from the defined address right after reset. For example, this boot configuration ca be used to execute a custom bootloader from a memory that is attached via the Processor-External Bus Interface (XBUS).

The Bootloader ROM (BOOTROM) is not enabled / implement at all. If the processor-internal Instruction Memory (IMEM) is enabled it will be implemented as blank RAM.

This configuration will implement the Instruction Memory (IMEM) as pre-initialized read-only memory (ROM). The ROM is initialized during synthesis with the according application image file (rtl/core/neorv32_application_image.vhd). After reset, the CPU will directly start executing this image. Since the IMEM is implemented as ROM, the executable cannot be altered at runtime at all.

The Bootloader ROM (BOOTROM) is not enabled / implement at all.

Overview

Implementation of the processor-internal instruction memory is enabled by the processor’s MEM_INT_IMEM_EN generic. The total memory size in bytes is defined via the MEM_INT_IMEM_SIZE generic. Note that this size should be a power of two to optimize physical implementation. If enabled, the IMEM is mapped to base address 0x00000000 (see section Address Space).

By default the IMEM is implemented as true RAM so the content can be modified during run time. This is required when using the Bootloader (or the [_on_chip_debugger]) so it can update the content of the IMEM at any time.

Alternatively, the IMEM can be implemented as pre-initialized read-only memory (ROM), so the processor can directly boot from it after reset. This option is configured via the BOOT_MODE_SELECT generic. See section Boot Configuration for more information. The initialization image is embedded into the bitstream during synthesis. The software framework provides an option to generate and override the default VHDL initialization file rtl/core/neorv32_application_image.vhd, which is automatically inserted into the IMEM (see Makefile Targets. If the IMEM is implemented as RAM (default), the memory block will not be initialized at all.

Overview

Implementation of the processor-internal data memory is enabled by the processor’s MEM_INT_DMEM_EN generic. The total memory size in bytes is defined via the MEM_INT_DMEM_SIZE generic. Note that this size should be a power of two to optimize physical implementation. If the DMEM is implemented, it is mapped to base address 0x80000000 by default (see section Address Space). The DMEM is always implemented as true RAM.

Overview

The boot ROM contains the executable image of the default NEORV32 Bootloader. When the Boot Configuration is set to bootloader mode (0) via the BOOT_MODE_SELECT generic, the boot ROM is automatically enabled and the CPU boot address is automatically adjusted to the base address of the boot ROM.

Overview

The processor features an optional instruction cache to improve performance when using memories with high access latencies. The cache is connected directly to the CPU’s instruction fetch interface and provides full-transparent accesses. The cache is direct-mapped and read-only.

Overview

The processor features an optional data cache to improve performance when using memories with high access latencies. The cache is connected directly to the CPU’s data access interface and provides full-transparent accesses. The cache is direct-mapped and uses "write-allocate" and "write-back" strategies.

Overview

The NEORV32 DMA provides a small-scale scatter/gather direct memory access controller that allows to transfer and modify data independently of the CPU. A single read/write transfer channel is implemented that is configured via memory-mapped registers. a configured transfer can either be triggered manually or by a programmable CPU FIRQ interrupt (see NEORV32-Specific Fast Interrupt Requests).

The DMA is connected to the central processor-internal bus system (see section Address Space) and can access the same address space as the CPU core. It uses interleaving mode accessing the central processor bus only if the CPU does not currently request and bus access.

The controller can handle different data quantities (e.g. read bytes and write them back as sign-extend words) and can also change the Endianness of data while transferring.

Theory of Operation

The DMA provides four memory-mapped interface registers: A status and control register CTRL and three registers for configuring the actual DMA transfer. The base address of the source data is programmed via the SRC_BASE register. Vice versa, the base address of the destination data is programmed via the DST_BASE. The third configuration register TTYPE is use to configure the actual transfer type and the number of elements to transfer.

The DMA is enabled by setting the DMA_CTRL_EN bit of the control register. Manual trigger mode (i.e. the DMA transfer is triggered by writing to the TTYPE register) is selected if DMA_CTRL_AUTO is cleared. Alternatively, the DMA transfer can be triggered by a processor internal FIRQ signal if DMA_CTRL_AUTO is set (see section below).

The DMA uses a load-modify-write data transfer process. Data is read from the bus system, internally modified and then written back to the bus system. This combination is implemented as an atomic progress, so canceling the current transfer by clearing the DMA_CTRL_EN bit will stop the DMA right after the current load-modify-write operation.

If the DMA controller detects a bus error during operation, it will set either the DMA_CTRL_ERROR_RD (error during last read access) or DMA_CTRL_ERROR_WR (error during last write access) and will terminate the current transfer. Software can read the SRC_BASE or DST_BASE register to retrieve the address that caused the according error. Alternatively, software can read back the NUM bits of the control register to determine the index of the element that caused the error. The error bits are automatically cleared when starting a new transfer.

When the DMA_CTRL_DONE flag is set the DMA has actually executed a transfer. However, the DMA_CTRL_ERROR_* flags should also be checked to verify that the executed transfer completed without errors. The DMA_CTRL_DONE flag is automatically cleared when writing the CTRL register.

Transfer Configuration

If the DMA is set to manual trigger mode (DMA_CTRL_AUTO = 0) writing the TTRIG register will start the programmed DMA transfer. Once started, the DMA will read one data quantity from the source address, processes it internally and then will write it back to the destination address. The DMA_TTYPE_NUM bits of the TTYPE register define how many times this process is repeated by specifying the number of elements to transfer.

Optionally, the source and/or destination addresses can be increments according to the data quantities automatically by setting the according DMA_TTYPE_SRC_INC and/or DMA_TTYPE_DST_INC bit.

Four different transfer quantities are available, which are configured via the DMA_TTYPE_QSEL bits:

Optionally, the DMA controller can automatically convert Endianness of the transferred data if the DMA_TTYPE_ENDIAN bit is set.

Automatic Trigger

As an alternative to the manual trigger mode, the DMA can be set to automatic trigger mode starting a pre-configured transfer if a specific processor-internal peripheral issues a FIRQ interrupt request. The automatic trigger mode is enabled by setting the CTRL register’s DMA_CTRL_AUTO bit. In this configuration no transfer is started when writing to the DMA’s TTYPE register.

The actually triggering FIRQ channel is configured via the control register’s DMA_CTRL_FIRQ_SEL bits. Writing a 0 will select FIRQ channel 0, writing a 1 will select FIRQ channel 1, and so on. See section Processor Interrupts for a list of all FIRQ channels and their according sources.

The FIRQ trigger can operate in two trigger mode configured via the DMA_CTRL_FIRQ_TYPE flag:

Memory Barrier / Fence Operation

Optionally, the DMA can issue a FENCE request to the downstream memory system when a transfer has been completed without errors. This can be used to re-sync caches (flush and reload) and buffers to maintain data coherency. This automatic fencing is enabled by the setting the control register’s DMA_CTRL_FENCE bit.

DMA Interrupt

The DMA features a single CPU interrupt that is triggered when the programmed transfer has completed. This interrupt is also triggered if the DMA encounters a bus error during operation. The interrupt will remain pending until the control register’s DMA_CTRL_DONE is cleared (this will happen upon any write access to the control register).

Register Map

Overview

The external bus interface provides a Wishbone b4-compatible on-chip bus interface that is implemented if the XBUS_EN generic is true. This bus interface can be used to attach processor-external modules like memories, custom hardware accelerators or additional peripheral devices. An optional cache module ("XCACHE") can be enabled to improve memory access latency.

Wishbone Bus Protocol

The external bus interface complies to the pipelined Wishbone b4 protocol. Even though this protocol was explicitly designed to support pipelined transfers, only a single transfer will be "in fly" at once. Hence, just two types of bus transactions are generated by the XBUS controller (see images below).

An accessed XBUS/Wishbone device does not have to respond immediately to a bus request by sending an ACK. Instead, there is a time window where the device has to acknowledge the transfer. This time window is configured by the XBUS_TIMEOUT generic and it defines the maximum time (in clock cycles) a bus access can be pending before it is automatically terminated raising an bus fault exception. If XBUS_TIMEOUT is set to zero, the timeout is disabled and a bus access can take an arbitrary number of cycles to complete. Note that this is not recommended as a missing ACK will permanently stall the entire processor!

Furthermore, an accesses XBUS/Wishbone device can signal an error condition at any time by setting the ERR signal high for one cycle. This will also terminate the current bus transaction before raising a CPU bus fault exception.

Access Tag

The XBUS tag signal xbus_tag_o(0) provides additional information about the current access cycle. It compatible to the the AXI4 ARPROT and AWPROT signals.

External Bus Cache (X-CACHE)

The XBUS interface provides an optional internal cache that can be used to buffer processor-external accesses. The x-cache is enabled via the XBUS_CACHE_EN generic. The total size of the cache is split into the number of cache lines or cache blocks (XBUS_CACHE_NUM_BLOCKS generic) and the line or block size in bytes (XBUS_CACHE_BLOCK_SIZE generic).

The cache uses a direct-mapped architecture that implements "write-allocate" and "write-back" strategies. The write-allocate strategy will fetch the entire referenced block from main memory when encountering a cache write-miss. The write-back strategy will gather all writes locally inside the cache until the according cache block is about to be replaced. In this case, the entire modified cache block is written back to main memory.

Overview

The stream link interface provides independent RX and TX channels for sending and receiving stream data. Each channel features a configurable internal FIFO to buffer stream data (IO_SLINK_RX_FIFO for the RX FIFO, IO_SLINK_TX_FIFO for the TX FIFO). The SLINK interface provides higher bandwidth and less latency than the external bus interface making it ideally suited for coupling custom stream processors or streaming peripherals.

Interface & Protocol

The SLINK interface consists of four signals for each channel:

Theory of Operation

The SLINK provides four interface registers. The control register (CTRL) is used to configure the module and to check its status. Two individual data registers (DATA and DATA_LAST) are used to send and receive the link’s actual data stream.

The DATA register provides direct access to the RX/TX FIFO buffers. Read accesses return data from the RX FIFO. After reading data from this register the control register’s SLINK_CTRL_RX_LAST flag can be checked to determine if the according data word has been marked as "end of stream" via the slink_rx_lst_i signal (this signal is also buffered by the link’s FIFO). Writing to the DATA register will immediately write to the TX link FIFO. When writing to the TX_DATA_LAST the according data word will also be marked as "end of stream" via the slink_tx_lst_o signal (this signal is also buffered by the link’s FIFO).

The configured FIFO sizes can be retrieved by software via the control register’s SLINK_CTRL_RX_FIFO_* and SLINK_CTRL_TX_FIFO_* bits.

The SLINK is globally activated by setting the control register’s enable bit SLINK_CTRL_EN. Clearing this bit will reset all internal logic and will also clear both FIFOs. The FIFOs can also be cleared manually at any time by setting the SLINK_CTRL_RX_CLR and/or SLINK_CTRL_TX_CLR bits (these bits will auto-clear).

The current status of the RX and TX FIFOs can be determined via the control register’s SLINK_CTRL_RX_* and SLINK_CTRL_TX_* flags.

Stream Routing Information

Both stream link interface provide an optional port for routing information: slink_tx_dst_o (AXI stream’s TDEST) can be used to set a destination address when using a switch/interconnect to access several stream sinks. slink_rx_src_i (AXI stream’s TID) can be used to determine the source when several sources can send data via a switch/interconnect. The routing information can be set/read via the ROUTE interface registers. Note that all routing information is also fully buffered by the internal RX/TX FIFOs. RX routing information has to be read after reading the according RX data. Vice versa, TX routing information has to be set before writing the according TX data.

Interrupts

The SLINK module provides two independent interrupt channels: one for RX events and one for TX events. The interrupt conditions are based on the according link’s FIFO status flags and are configured via the control register’s SLINK_CTRL_IRQ_* flags. The according interrupt will fire when the module is enabled (SLINK_CTRL_EN) and the selected interrupt conditions are met. Note that all enabled interrupt conditions are logically OR-ed per channel. If any enable interrupt conditions becomes active the interrupt will become pending until the interrupt-causing condition is resolved (e.g. by reading from the RX FIFO).

Register Map

Overview

The general purpose parallel IO unit provides a simple parallel input and output port. These ports can be used chip-externally (for example to drive status LEDs, connect buttons, etc.) or chip-internally to provide control signals for other IP modules.

The actual number of input/output pairs is defined by the IO_GPIO_NUM generic. When set to zero, the GPIO module is excluded from synthesis and the output port gpio_o is tied to all-zero. If IO_GPIO_NUM is less than the maximum value of 64, only the LSB-aligned bits in gpio_o and gpio_i are actually connected while the remaining bits are tied to zero or are left unconnected, respectively.

Register Map

Overview

The cyclic redundancy check unit provides a programmable checksum computation module. The unit operates on single bytes and can either compute CRC8, CRC16 or CRC32 checksums based on an arbitrary polynomial and start value.

Theory of Operation

The module provides four interface registers:

The MODE, POLY and SREG registers need to be programmed before the actual processing can be started. Writing a byte to DATA will update the current checksum in SREG.

Register Map

Overview

The watchdog (WDT) provides a last resort for safety-critical applications. When a pre-programmed timeout value is reached a system-wide hardware reset is generated. The internal counter has to be reset explicitly by the application program every now and then to prevent a timeout.

Theory of Operation

The watchdog is enabled by setting the control register’s WDT_CTRL_EN bit. When this bit is cleared, the internal timeout counter is reset to zero and no system reset can be triggered by this module.

The internal 32-bit timeout counter is clocked at 1/4096th of the processor’s main clock (f_WDT[Hz] = f_main[Hz] / 4096). Whenever this counter reaches the programmed timeout value (WDT_CTRL_TIMEOUT bits in the control register) a hardware reset is triggered.

The watchdog’s timeout counter is reset ("feeding the watchdog") by writing the reset PASSWORD to the RESET register. The password is hardwired to hexadecimal 0x709D1AB3.

Configuration Lock

The watchdog control register can be locked to protect the current configuration from being modified. The lock is activated by setting the WDT_CTRL_LOCK bit. In the locked state any write access to the control register is entirely ignored (see table below, "writable if locked"). However, read accesses to the control register as well as watchdog resets are further possible.

The lock bit can only be set if the WDT is already enabled (WDT_CTRL_EN is set). Furthermore, the lock bit can only be cleared again by a system-wide hardware reset.

Strict Mode

The strict operation mode provides additional safety functions. If the strict mode is enabled by the WDT_CTRL_STRICT control register bit an immediate hardware reset if enforced if

Cause of last Hardware Reset

The cause of the last system hardware reset can be determined via the WDT_CTRL_RCAUSE_* bits:

Register Map

Overview

The MTIME module implements a memory-mapped machine system timer that is compatible to the RISC-V privileged specifications. The 64-bit system time is accessed via individual TIME_LO and TIME_HI registers. A 64-bit time compare register, which is accessible via individual TIMECMP_LO and TIMECMP_HI registers, can be used to configure the CPU’s machine timer interrupt (MTI)).

The interrupt is triggered whenever TIME (high & low part) is greater than or equal to TIMECMP (high & low part). The interrupt remains active (=pending) until TIME becomes less than TIMECMP again (either by modifying TIME or TIMECMP). The current system time is available for other SoC modules via the top’s mtime_time_o signal.

Register Map

Overview

The NEORV32 UART provides a standard serial interface with independent transmitter and receiver channels, each equipped with a configurable FIFO. The transmission frame is fixed to 8N1: 8 data bits, no parity bit, 1 stop bit. The actual transmission rate (Baud rate) is programmable via software. The module features two memory-mapped registers: CTRL and DATA. These are used for configuration, status check and data transfer.

RX and TX FIFOs

The UART provides individual data FIFOs for RX and TX to allow data transmission without CPU intervention. The sizes of these FIFOs can be configured via the according configuration generics (UART0_RX_FIFO and UART0_TX_FIFO). Both FIFOs a re automatically cleared when disabling the module via the UART_CTRL_EN flag. However, the FIFOs can also be cleared individually by setting the UART_CTRL_RX_CLR / UART_CTRL_TX_CLR flags.

Theory of Operation

The module is enabled by setting the UART_CTRL_EN bit in the UART0 control register CTRL. The Baud rate is configured via a 10-bit UART_CTRL_BAUDx baud divisor (baud_div) and a 3-bit UART_CTRL_PRSCx clock prescaler (clock_prescaler).

Baud rate = (f_main[Hz] / clock_prescaler) / (baud_div + 1)

The control register’s UART_CTRL_RX_* and UART_CTRL_TX_* flags provide information about the RX and TX FIFO fill level. Disabling the module via the UART_CTRL_EN bit will also clear these FIFOs.

A new TX transmission is started by writing to the DATA register. The transfer is completed when the UART_CTRL_TX_BUSY control register flag returns to zero. RX data is available when the UART_CTRL_RX_NEMPTY flag becomes set. The UART_CTRL_RX_OVER will be set if the RX FIFO overflows. This flag is cleared only by disabling the module via UART_CTRL_EN.

UART Interrupts

The UART module provides independent interrupt channels for RX and TX. These interrupts are triggered by certain RX and TX FIFO levels. The actual configuration is programmed independently for the RX and TX interrupt channel via the control register’s UART_CTRL_IRQ_RX_* and UART_CTRL_IRQ_TX_* bits:

Once an UART interrupt has fired it remains pending until the actual cause of the interrupt is resolved; for example if just the UART_CTRL_IRQ_RX_NEMPTY bit is set, the RX interrupt will keep firing until the RX FIFO is empty again.

RTS/CTS Hardware Flow Control

The NEORV32 UART supports optional hardware flow control using the standard CTS uart0_cts_i ("clear to send") and RTS uart0_rts_o ("ready to send" / "ready to receive (RTR)") signals. Both signals are low-active. Hardware flow control is enabled by setting the UART_CTRL_HWFC_EN bit in the modules control register CTRL.

When hardware flow control is enabled:

Simulation Mode

The UART provides a simulation-only mode to dump console data as well as raw data directly to a file. When the simulation mode is enabled (by setting the UART_CTRL_SIM_MODE bit) there will be no physical transaction on the uart0_txd_o signal. Instead, all data written to the DATA register is immediately dumped to a file. Data written to DATA[7:0] will be dumped as ASCII chars to a file named neorv32.uart0_sim_mode.out. Additionally, the ASCII data is printed to the simulator console.

Both file are created in the simulation’s home folder.

Register Map